Shows the time taken for generating a single reply from a Japanese LLM fine tuned on a suitable instruction dataset. Shows the model name, reply time and number of replies generated in an hour on an Nvidia H100 SXM GPU.